Business Intelligence in the Era of Big Data and Cognitive Business

Business Intelligence in the Era of Big Data and Cognitive Business

CHAPTER 2 Business Intelligence in the Era of Big Data and Cognitive Business Business executives, managers, and analysts have wrestled for over two...

371KB Sizes 0 Downloads 51 Views

CHAPTER

2

Business Intelligence in the Era of Big Data and Cognitive Business Business executives, managers, and analysts have wrestled for over two decades with the problem of understanding how to leverage data to improve business results. For much of that time, the umbrella term “business intelligence”—or “BI” for short—has been used to describe a family of business analysis techniques ranging from standard reports to highly sophisticated advanced statistics. More recently, terms like “big data” and “cognitive business” have been introduced into the business and technical lexicon. Upon close examination, the newer terminology is about the same thing that BI has always been about: analyzing the vast amounts of data that companies generate and/or purchase in the course of business as a means of improving profitability and competitiveness. Accordingly, we will use the terms BI and business intelligence throughout the book, and we will discuss the newer concepts as appropriate. Whether we call it BI, data mining, big data, cognitive business or whatever, the business challenges for realizing business value are the same: 1. helping business executives, managers, and analysts in companies sort through the confusing array of terminology to understand what is real, what is hype, and how to leverage data throughout the enterprise to improve business results; 2. ensuring alignment between business strategies, the core business processes that execute the strategies, and the use of BI to improve those core business processes—processes such as marketing, sales, customer service, and operations that ultimately determine the economic results for the business; and 3. managing the complex organizational factors that determine how effectively BI applications are developed and how effectively they are adopted within business processes to increase revenues, reduce costs, or both. Business Intelligence Strategy and Big Data Analytics. DOI: http://dx.doi.org/10.1016/B978-0-12-809198-2.00002-6 © 2016 Elsevier Inc. All rights reserved.

28

Business Intelligence Strategy and Big Data Analytics

Because of struggles in these regards, BI is underutilized within companies where it could have a substantial impact—at a time when information and analysis have become critical factors in business success. A key aspect of the problem has to do with lack of clarity about what BI is, what BI they should have, and how BI is related to analytics, big data, data warehousing, and other related topics. With this in mind, it is important to clarify some of the terminology we will use throughout the book. We’ll also explore what BI success means to different people in different industries and job functions.

2.1 GETTING CLEAR ABOUT TERMINOLOGY—BUSINESS DEFINITIONS OF BUSINESS INTELLIGENCE AND RELATED TERMS Having worked with business executives, managers, and analysts in well-known companies in a wide range of industries, I can say with certainty that they are often unclear about both the terminology and the value propositions associated with BI. It’s no surprise given the confusing array of BI-related terminology to which they are exposed, as exemplified by Fig. 2.11. With this in mind, we’ll use the following business-oriented definitions throughout the book2: • Business intelligence (BI): An umbrella term that encompasses provision of relevant reports, scorecards, dashboards, e-mail alerts, prestructured user-specified queries, ad hoc query capabilities, multidimensional analyses, statistical analyses, forecasts, models, and/or simulations to business users for use in increasing revenues, reducing costs, or both. • Analytics: A subset of BI and an umbrella term that encompasses provision of relevant statistical analyses, forecasts, models, and/or simulations to business users for use in increasing revenues, reducing costs, or both. • Big data: Large amounts of rapidly generated pictures, video clips, location (geospatial) data, sensor data, text messages, document images, web logs, and machine data traditionally captured and used 1 From Williams S. 5 Barriers to BI Success and how to overcome them. Strategic Finance, July 2011. 2 Adapted from Williams S. Big data strategy approaches: business-driven or discovery-based? Bus Intellig J 19(4).

Business Intelligence in the Era of Big Data and Cognitive Business

29

Some of the many sides of business intelligence Styles of BI

New age BI

New age data

Functional analytics

Reporting

Agile BI

Social media data

Performance management analytics

Ad hoc query

SaaS BI

Unstructured data

Parameterized queries

Pervasive BI & analytics

Mobile data

OLAP

Cognitive business

Big data

Advanced analytics

Self-service analytics

Sensor data

Predictive analytics

Social analytics

Machine data

Cognitive business techniques

Tools Scorecarding & dashboarding

Financial analytics Supply chain analytics Customer analytics Operations analytics HR analytics

Real-time analytics Mobile BI and analytics

Query & reporting Statistics & data mining OLAP cubes ETL Master data management Metadata management

“Fill in the blank” analytics

Data governance

Figure 2.1 BI terminology can be confusing for those who don’t work with it day-in and day-out.









by social media and Internet-based businesses and more recently being leveraged by early adopter mainstream businesses. Big data analytics: Analysis of stored big data content of various kinds to supplement BI and traditional analytics for use in increasing revenues, reducing costs, or both. Also useful for nonbusiness uses, such as public safety and national defense. Structured data: The typical business data used by companies for decades—represented as numerical values, calculated measures and metrics, and business facts such as financial results, customer characteristics, factory output, or product characteristics—and which has been typically stored in relational databases. Unstructured data: Digital content such as pictures, video clips, text messages, document images, and web logs. “Unstructured data” is substantially equivalent to “big data”—but differs in that sensor data, location data, and machine data are typically structured data and are included as examples of the variety of data that collectively constitute big data. Cognitive business: The use of structured and unstructured data and highly sophisticated analytical techniques to identify, evaluate,

30

Business Intelligence Strategy and Big Data Analytics

and recommend business courses of actions. Related terms include artificial intelligence and machine learning. • Data warehouse: A specialized database used to store important business information about transactions, products, customers, channels, financial results, performance metrics, and other business information over multiple years so that the business information can be easily and consistently used to improve business results. As we proceed through the forthcoming chapters, we’ll use the terms “business intelligence” and “BI” in the broad sense defined above, that is, as an umbrella term. We will also use it to encompass the newer concepts—big data, big data analytics, and cognitive business. Where appropriate to the context, we will distinguish between BI in general and a specific type of BI—such as analytics or multidimensional analysis.

2.2 THE HYPE AROUND BI, BIG DATA, ANALYTICS, AND COGNITIVE BUSINESS Every day, executives and managers at leading companies are bombarded with claims about BI, big data, analytics, and cognitive business. Many business people are a skeptical lot when it comes to potential business improvements enabled by information technology. They need to have a concrete idea of how BI, analytics, cognitive business, and/or big data would actually help them in their specific business before approving multimillion dollar budgets. As one client put it, “we need to sort through what is hype and what is real for our specific business context.” Prompted by this need, I suggest the following considerations. It might be hype if. . . the smartest, most experienced business people at your company cannot explain very specifically how having better information and analyses would enable the company to capture incremental revenues and/or reduce expenses. In our professional opinion, the “true North” by which to navigate the hype is whether or not there is a clear, concrete connection between a proposed use of BI and an important company business process that makes a difference to customers and company economics. In this case, “better information” typically means transactional history, product/service holdings, plus customer demographic information about each and every individual customer, automatically available on a daily basis, and organized for reuse across the company on a daily/weekly/monthly basis

Business Intelligence in the Era of Big Data and Cognitive Business

31

whenever such information is needed to run the company and improve profits. Unless that connection can be made in a very specific and detailed way, it might be hype. It might be hype if. . . there is a big gap between the visionary and poetic language being used to describe benefits of a BI, big data, or cognitive business product offering and the actual products and services being sold. We worked with a $2 1 billion company in 2011/12 that was trying to figure out what to do in the BI and analytics space so they asked leading BI vendors what they should do. One prominent vendor known for flashy ads aimed at business executives submitted a proposal that: • was mostly about selling licenses for commodity BI tools that have been on the market for over a decade and • hoped to sell some 2000 full-featured BI tool licenses to a company that was not likely to need that many licenses for years to come, if ever. The business benefits being touted were couched in high-level business terms like agile business and customer intimacy, but what was being sold was a package of canned reports with little connection to the business benefits being claimed. If you perceive this kind of gap, it might be hype. It might be hype if. . . you’re being sold a race car and your company is just learning to ride a bike. When it comes to BI, big data, analytics, and cognitive business, even the most successful companies in many industries are just starting to move up the maturity curve. One of our clients was being pushed by a leading vendor to purchase roughly $500,000 in advanced enterprise analytics hardware and software, when what the client really needed to get started was two desktop licenses for a standard statistical analysis package—for a total of roughly $24,000. If you feel you are being sold a package that represents what your company might need after several years of getting its feet wet, it might be hype. It might be hype if. . . the topic is big data, big data analytics, or cognitive business. The “next big thing” in BI comes along every couple of years and then fizzles. Pushed by big consulting firms, big vendors, and prominent analyst firms, these latest concepts are all the rage at a given point in time. Now of course the proponents point to case studies that back up their claims, but many cases are really just about

32

Business Intelligence Strategy and Big Data Analytics

creative uses of BI and analytics using traditional data that we’ve had for decades. Our experience with traditional Fortune 1000 and midcap companies is that: 1. many successful big companies haven’t even leveraged regular data yet, let alone big data or cognitive computing; 2. the most valuable data for BI and analytics are generally the common, mundane transactional data, customer data, and financial data that companies have had for years—data that is the key to understanding the economic performance of the company and what drives it; 3. big data in the form of unstructured digital content such as pictures, video clips, text messages, and document images is of unproven value in many traditional (non-Internet based) companies; and 4. many traditional companies do not generate unstructured digital content in the normal course of their business, though marketers are starting to leverage web and social media data. If the value of big data is not clear to the smartest, most experienced people in your company, it might be hype. It might be hype if. . . you hear the term “out of the box” in relation to any BI, analytics, big data, or cognitive business software product or service. Software vendors design standard products that they hope to license to millions of users. While there is some tailoring of the products to industries and/or job functions, these products are simply prepackaged capabilities that have the potential to help companies leverage business information and analytics to create business value. That “out of the box” potential means nothing without intelligent use of the potential to create incremental revenues and/or reduce or optimize expenses. Vendors are sophisticated at convincing business executives and managers that their “solution” reduces risk, speeds up time to value, and creates competitive advantage—out of the box and without any customization. In other words, their product is a silver bullet for solving all manner of complicated business challenges. If this sounds too good to be true, it might be hype. It might be hype if. . . a technology vendor conveys the idea that all one needs to do is buy their product and the company will obtain benefits like improved profits. BI, analytics, big data, and cognitive business need to be business-driven initiatives, not technology-driven. If a

Business Intelligence in the Era of Big Data and Cognitive Business

33

company can develop a clear vision and concrete strategy for leveraging business information and business analyses, the technology piece can be figured out and is generally low-risk if one uses tools that have been around for a long time. There is no substitute for aligning BI, analytics, big data, and/or cognitive business applications with core business process, managing process and cultural change, and driving adoption of the applications by business users. If a vendor claims that the technology delivers the BI benefits, it might be hype. As we noted at the outset, there is a lot of hype around BI, analytics, big data, and cognitive business. This makes it hard for business professionals to understand their true opportunities, understand the risks, and formulate pragmatic strategies and program plans. We hope this book will paint a picture of what is possible with BI and what may make sense for your industry, company, and job function. Armed with this information, you’ll be in a stronger position to sort through the hype. And if your company has been paying BI consultants and/or BI vendors, you’ll be in a better position to judge whether your company is better off for having done so.

2.3 A BUSINESS VIEW OF BIG DATA3 From a business perspective, what’s really new and important about big data? The most widely communicated concept of big data holds that it differs from traditional data in its volume, variety, and velocity. Let’s examine those in turn. Data volume. There is no argument that the Internet and the social media revolution have spawned vast amounts of new kinds of data. And new technical approaches to storing and managing these vast volumes of new data have evolved to make the cost of keeping that data much less expensive. So we can store big data cheaply, but the “garbage in, garbage out” maxim still applies. From a business perspective, what is important is determining the utility of that data for creating business value. Data velocity. A good example of the change in velocity of data is provided by the electric utility industry and its adoption of smart meters. A utility with 700,000 customers might have obtained 700,000 3

Ibid.

34

Business Intelligence Strategy and Big Data Analytics

meter readings a month in the past. With smart meters, that same utility might obtain 700,000 meter readings a minute. More broadly, the explosion of social media activity and Internet commerce means that there are hundreds of millions of pieces of data created every second. From a business perspective, what is important is determining whether and how high data velocities are relevant and useful for creating business value. Data variety. What is supposedly new with big data is the capture and storage of unstructured data or semistructured data—all of it digital but much of it not really “data” in the traditional sense of the word. The pictures, video clips, text messages, document images, and web logs stored today could arguably be called “content” or “digital content” rather than data. In fact, in the document management and workflow worlds, many of these types of unstructured data are considered content. Insurance companies for years have captured pictures and copies of documents and stored them within workflow-oriented claims processing systems. In the banking world, the Check 21 initiative was based on storing check images on optical disks. In the manufacturing world, statistical process control methods that generate large volumes of sensor readings have been in use for decades. From a business perspective, what is important is determining how these various forms of “big data content” can be used to create business value. Based on the above, it seems fair to conclude that the volume and velocity of digital content creation is indeed new, and that there are new varieties of digital content—with text messages and web logs (less new) being good examples. As to the business importance of big data, we might reasonably point out that: • new varieties of digital content will be important if they can be used to increase revenues, reduce costs, or both—and this will depend on industry-specific and company-specific factors; • increased volumes of digital content will be important if the content can be used to increase revenues, reduce costs, or both—otherwise one could be spending money to store ever-increasing volumes of trash; and • increased velocities of digital content will be important if the content can be used to increase revenues, reduce costs, or both and its utility for doing so is time-dependent—otherwise one could be spending money to accumulate trash more quickly.

Business Intelligence in the Era of Big Data and Cognitive Business

35

With all of the foregoing in mind, and giving due recognition to the fact that big data might be valuable, the potential competitive implications of big data suggest that companies systematically evaluate big data opportunities when formulating their BI strategies—or as extensions to existing BI strategies. Such an evaluation should consider: 1. alignment with any ongoing enterprise, business unit, or functional uses of BI; 2. how big data content—pictures, video clips, location (geospatial) data, sensor data, text messages, document images, web logs, and machine data—can be used to increase revenues, reduce costs, or both; and 3. whether to invest in capturing and storing big data content “on the come”—by which I mean ahead of any clear idea of exactly how that content will be used to increase revenues, reduce costs, or both. There is nearly 20 years of history of companies using BI, data warehouses, and traditional analytics to create business value—with many successful companies still needing to do more to fully leverage these proven tools. And there are proven methods that can be applied for analyzing how big data content can be used to create business value—about which we will say more in chapter “The Strategic Importance of Business Intelligence.” From a BI perspective, big data is simply another potential source of useful information and digital content that might be useful for analytical purposes aimed at improving the business processes that drive economic results.

2.4 A BUSINESS VIEW OF COGNITIVE BUSINESS The field of cognitive science draws on disciplines such as neuroscience, psychology, artificial intelligence, statistics, mathematics, and computer science. Cognitive business is simply the use of cognitive science techniques and methods to address complex, dynamic, and/or ambiguous business situations. Since many business situations and decision contexts have those characteristics, the thinking is that using cognitive science for business purposes can result in better business performance than would otherwise be possible. If we look at the idea of cognitive business from a BI perspective, we can compare the two as shown in Table 2.1.

36

Business Intelligence Strategy and Big Data Analytics

Table 2.1 Cognitive Business Techniques Add New Analytical and Decision Support Tools to the BI Toolkit Business Intelligence

Cognitive Business

Business outputs Reports

X

Scorecards and dashboards

X

Multidimensional analysis

X

Ad hoc analysis

X

Advanced analytics

X

X

Predictive analytics

X

X

Alerts

X

Visualization

X

Relationship to business processes Information and analysis about process performance

X

Information and analysis used within processes

X

X

Information and analysis for process control

X

X

X

X

Data inputs Traditional structured data Unstructured data

X

Platform Company premises

X

X

Cloud based

X

X

The circled portions of Table 2.1 highlight similarities and differences as follows: 1. Similarity—both BI and cognitive business encompass the use of standard mathematical and statistical methods to perform analyses in the context of various business domains. For example, advanced analytics (backward looking, such as trend analysis) and predictive analytics (forward looking, such as simulations and optimizations) have been considered a subset of BI for decades. Cognitive business is mainly about such analytical methods, although it may bring more advanced techniques to bear than have traditionally been used in business. Arguably, any analytical technique employed by cognitive business applications and applied to structured data can also be used

Business Intelligence in the Era of Big Data and Cognitive Business

37

by data scientists and business domain experts working with analytics platforms such as SAS and SPSS. 2. Difference—traditional BI tools are used to analyze structured data, whereas cognitive business applications would have both structured and unstructured data as inputs. Given the need to analyze unstructured data, there are a variety of tools that are used to basically take unstructured data and describe it in ways so that it can be processed by computerized algorithms. From a business perspective, what might be important is if your particular company needs to move beyond more traditional BI and analytics to analyze large amounts of unstructured data in order to improve some relevant business process in a way that increases revenues, reduces costs, or both. In an industry where there are such needs, it would also be important to consider the competitive implications of cognitive business techniques. From a strategy perspective, there is a need to evaluate cognitive business applications developed by vendors versus building a customized application. In the latter case, a company’s internal domain experts would work with data scientists and leverage packaged analytical components (eg, text analysis software) to weave together cognitive business algorithms. From a BI strategy perspective, from this point forward we will consider cognitive business applications to be a type of BI—an extension of traditional analytics and a subset of BI.

2.5 BI AND ANALYTICS—IS THERE A DIFFERENCE? For our purposes, business analytics are data-based applications of quantitative analysis methods in use in businesses for decades. There are hundreds of books that apply various quantitative analysis, operations research, and discrete mathematics methods to specific business domains, ranging from sophisticated customer segmentations and predictions of customer lifetime value to demand forecasting and supply chain optimization. So analytics, per se, are not new. Rather, proven quantitative analysis methods have been implemented as packaged software applications and bundled into “analytics platforms” that are used to build a wide range of analytical applications that address common business challenges. SAS and SPSS are well-known examples of companies that sell analytics platforms, and there are many others.4 4 This paragraph is excerpted from Williams S. Analytics: a tool executives and managers need to embrace. MWorld, J Am Manage Assoc, Winter 2012 2013.

38

Business Intelligence Strategy and Big Data Analytics

More broadly, we previously defined BI as an umbrella term that encompasses provision of relevant reports, scorecards, dashboards, email alerts, prestructured user-specified queries, ad hoc query capabilities, multidimensional analyses, statistical analyses, forecasts, models, and/or simulations to business users for use in increasing revenues, reducing costs, or both. Typical business intelligence (BI) applications—all of which leverage business data and provide analytical perspectives— include: • REPORTS: standard, preformatted information for backwardlooking analysis of business trends, events, and performance results; • MULTIDIMENSIONAL ANALYSES: applications that leverage a common database of trusted business information and that fully automate information slicing and dicing for analysis of the underlying drivers of business events, trends, and performance results; • SCORECARDS and DASHBOARDS: convenient forms of multidimensional analyses that are common across an organization, that enable rapid evaluation of business trends, events, and performance results, and that facilitate use of a common management framework and vocabulary for measuring, monitoring, and improving business performance; • ADVANCED ANALYTICS: automated applications that distill historical business information so that past business trends, events, and results can be summarized and analyzed via well-known and long-used statistical methods; • PREDICTIVE ANALYTICS: automated applications that leverage historical business information, descriptive statistics, and/or stated business assumptions to predict or simulate future business outcomes that can be analyzed for their business impact; and • ALERTS: automated process control applications that analyze performance variables, compare results to a standard, and report variances outside defined performance thresholds. Ultimately, all of these forms of BI deliver business information for decision-makers to use to analyze past performance and its root causes, model and analyze various courses of actions, predict future results and analyze economic impacts, and make decisions that are informed by underlying data and sound analytical techniques. From this business perspective, advanced analytics and predictive analytics are a subset of BI, and the various kinds of BI leverage business information and analyses

Business Intelligence in the Era of Big Data and Cognitive Business

39

to inform decisions and drive business results. When it comes to formulating a BI strategy, companies should consider all forms of BI—including analytics. When it comes to defining BI requirements, it is important to be clear about what type of BI is needed, for example, a scorecard or a report or an analytical application such as a forecast or a sophisticated customer segmentation application. In the end, BI has always been about analysis, and much of the “buzz” in the BI field these days is around a narrower conception of analytics—one centered on advanced analytics, predictive analytics, and big data analytics.

2.6 BEYOND THE HYPE—WHAT BI SUCCESS LOOKS LIKE BI is used to create business value by enabling increased revenues, reduced costs, or both—thus leading to increased profits. It is like a carpentry toolkit, where what needs to be built depends on the needs of the customer. With a carpentry toolkit, I can build a shed, a closet, a cabinet, a house, or whatever. With a standard BI toolkit, we can build custom BI applications that are designed to meet industryspecific and job-specific business challenges. Accordingly, BI success is a function of meeting those challenges. Success is demonstrated through improved business performance for the key business functions and processes of the firm, and thus it looks different to different executives and managers within different industries.

2.6.1 Industry Views of BI Success A central objective of BI is to provide executives, managers, and knowledge workers with information and analyses they can use to create positive business results. The information and analyses that are relevant in one industry may not be relevant in a different industry. For example, operations managers in a product distribution company are keenly interested in inventory levels, inventory turnover, and customer service trends because optimizing inventory in relation to customer service goals is critical to economic results. On the other hand, operations managers for a retail bank are primarily interested in serving customers quickly and cost-effectively and in offering additional products to customers based on what they are likely to need. BI for inventory analysis is critical for a distribution company, and is relatively less important for a retail bank. Because the uses of BI that are relevant differ by industry, BI success looks different depending on the industry in which a company operates. While not every company in an industry

40

Business Intelligence Strategy and Big Data Analytics

competes in the same way or has the same role in the value chain, we can still paint industry views of BI success using broad brushstrokes. For manufacturing companies, BI success consists of having the ability to actively manage and improve performance in the core areas that impact customer service and financial performance. A manufacturing company that has achieved a reasonable measure of BI success will have deployed the following BI applications: 1. Enterprise and business unit performance dashboards that are updated on a timely basis and that identify the unfavorable performance variances that require immediate management action— typically those variances related to revenues, product manufacturing costs and output, logistics performance, customer service, key supplier performance, and inventory. The variances displayed are typically for the top 10 or so contributors to the unfavorable variances, and the variances are calculated year-over-year, in relation to an annual operating plan or budget, in relation to updated operating plans or budgets, and by key dimensions such as customer, product, and channel. 2. Analytical dashboards that are reached from the performance dashboards and that allow managers and analysts to drill-down into the details of unfavorable variances so that corrective actions can be quickly identified, evaluated, decided upon, and acted upon. For example, if the performance dashboard identifies that two major customers are buying less and that product distribution in a targeted channel is below target, the analytical dashboard is the launching pad for identifying the root causes of the unfavorable variances. In our example, perhaps Customer A buys 10 products from us, and has decided to switch to a competitor’s product for two of the products. The analytical dashboard allows an analyst to see that Customer A is no longer ordering Products X and Y. Armed with this information, corrective action strategies are devised quickly and efficiently. 3. BI applications for enabling demand analysis and demand forecasting—typically a data mart with order and order line history and a combination of standard multidimensional analysis capabilities and an advanced and predictive analytics tool, such as SAS, SPSS, or a low-cost alternative. The demand analysis capability makes recurring processes, such as budgeting, sales and operations planning, setting inventory targets, establishing manufacturing plans,

Business Intelligence in the Era of Big Data and Cognitive Business

41

developing distribution plans, analyzing production capacity utilization, and developing brand/product plans and strategies, much more efficient. The demand forecasting capability is a key tool in company efforts to optimize costs, productivity, and asset utilization in relation to market and customer service requirements. The above BI applications enable manufacturers to actively measure, manage, control, and improve business performance in all the core business processes that determine customer service levels and revenue growth. They are high-level examples of the kinds of BI capabilities manufacturing companies require. Companies that have met these BI requirements can be said to have achieved BI success. For financial services companies, BI success typically consists of having the ability to offer personalized services and conduct intelligent, focused multichannel marketing campaigns that reach the right customers with the right offers and the right time. With possible exception of the investment banking and wealth management segments of the industry, financial services companies provide products and services that are generally commoditized, which means that competitive differentiation depends to a large degree on being able to offer differentiated customer service. Whether we’re taking about credit and debit cards, retail banking services, consumer lending, property and casualty insurance, or retirement and investment products, financial services companies face the challenge of treating large numbers of customers in a way that conveys that they know who they are and understand their individual needs. In this environment, financial services companies that have achieved a reasonable measure of BI success will have deployed the following BI applications: 1. BI applications that enable a so-called 360 view of each individual customer. This view provides basic information such as the customer’s name, address, and so forth. More importantly, it provides information about all aspects of the business relationship with the customer—such as account balances, loan balances, product/service holdings, loan payment history, credit and/or debit card transactions, deposits, withdrawals, and so forth. Also, it provides a complete record of all customer service and marketing interactions—including calls to a call center and their resolution and a record of all marketing offers made, which channel was used

42

Business Intelligence Strategy and Big Data Analytics

to make the offer, and whether the offer was accepted. This information is updated on a real-time or near real-time basis. 2. BI applications for multidimensional analysis of business performance by geography, market, location/office/branch, product, customer segment, and channel. Financial services companies are increasingly complex—with thousands or millions of customers, dozens of product variations, and an increasing number of digital interactions. Understanding how and where growth is being achieved, which products/services are doing well with which customers, and the trends in digital channel usage is fundamental to managing and improving customer service and business results. 3. BI applications for multidimensional analysis of marketing results and predicting customer propensity to purchase specific products and services. For years, financial services companies have used life stage and income as the primary bases for customer segmentation. The emergence of digital channels and the ability to mine call center records enables BI applications for more sophisticated segmentation, more personalized offers, more efficient list generation, and more real-time tracking of marketing campaign results. Further, the use of advanced and predictive analytics enables applications of segmentation based on predicted customer lifetime value and differentiated marketing and customer service tactics. The above BI applications enable financial services firms to cope with the inherent complexity of their business and to offer high-quality personalized customer service. They are high-level examples of the kinds of BI capabilities such companies require. Companies that have met these BI requirements can be said to have achieved BI success. For distributors, BI success typically consists of having the ability to effectively leverage information and analysis to manage margins, inventory levels, and customer service in a complex, dynamic, and lowmargin environment. While system distributors sometimes have a less complex environment, many distributors offer thousands of products to hundreds of customers who require delivery to thousands of endpoints. Product manufacturers offer a wide array of promotional deals, which distributors pass along in whole or in part to downstream distributors or retailers. The distributors themselves also offer deals—typically volume-based but also time-based and other variants. The net effect of this is that the distributor’s true product cost and true realized revenue

Business Intelligence in the Era of Big Data and Cognitive Business

43

on any given product is often unknown for weeks or months. This makes it hard to optimize pricing and promotions to achieve margin targets. This complexity is coupled with an often incomplete view of true demand at retail, which makes it difficult to understand price elasticity of demand and to optimize inventory in relation to customer service level requirements. In this environment, distributors that have achieved a reasonable measure of BI success will have deployed the following BI applications: 1. Executive and distribution center performance dashboards that are updated on a timely basis and that identify the unfavorable performance variances that require immediate management action— typically those variances related to revenues, product movement volumes, margins, inbound logistics performance, distribution center productivity, inventory levels, outbound logistics performance, product damage and returns, and customer service. The variances displayed are typically for the top 10 or so contributors to the unfavorable variances, and the variances are calculated year-over-year, in relation to an annual operating plan or budget, in relation to updated operating plans or budgets, and by key dimensions such as customer, product, and channel. 2. Analytical dashboards that are reached from the performance dashboards and that allow managers and analysts to drill-down into the details of unfavorable variances so that corrective actions can be quickly identified, evaluated, decided upon, and acted upon. For example, if the performance dashboard identifies product movement volume through a new channel is 30% below the targeted volume, the analytical dashboard is used to drill-down into the root causes of the variance. In this hypothetical, the analytical dashboard allows an analyst to see that the pricing used in an established channel was carried over to the new channel, and that has hindered product uptake. Armed with this information, corrective action strategies are devised quickly and efficiently. 3. BI applications for demand analysis and forecasting and for multidimensional analysis of marketing performance. Demand forecasting at the product and/or product family level is essential to optimizing purchasing quantities and inventory levels in relation to customer service requirements. Demand analysis is critical for optimizing pricing, promotions, and margins in relation to various demand scenarios. Multidimensional analysis of promotional performance provides

44

Business Intelligence Strategy and Big Data Analytics

critical insight into what promotion structures work best for which products, customers, channels, and geographic regions. The above BI applications enable distributors to cope with the inherent complexity of their business and to optimize margins, volume, inventory, and customer service. They are high-level examples of the kinds of BI capabilities such companies require. Companies that have met these BI requirements can be said to have achieved BI success. For utilities, BI success typically consists of having a robust and comprehensive set of system and plant operating performance information, detailed cost information, and customer service information—all of which allow the utility to meet targeted system reliability and customer service goals at costs that were assumed in rate justifications to regulatory bodies. Further, BI success for utilities includes extensive engineering information about assets—generation plants, substations, poles, underground wires, trucks, and so forth—for use in capital planning, project planning, maintenance planning, and predicting restoration times in responsive to outage events. Utilities operate in what amounts to a fixed-price environment where all constituencies want highly reliable supplies of electricity, natural gas, and water and very fast restoration times in the event of outages—all for low rates. In this environment, utilities that have achieved a reasonable measure of BI success will have deployed the following BI applications: 1. Executive and business unit performance dashboards that are updated on a timely basis and that identify the unfavorable performance variances that require immediate management action—typically those variances related to restoration times during outage events, system reliability, customer service, preventative maintenance progress, construction progress, energy costs, and safety performance. The variances displayed are typically for the top 10 or so contributors to the unfavorable variances, and the variances are calculated in relation to an annual operating plan or budget, in relation to updated operating plans or budgets, and by key dimensions such as customer type, power generation plant, geographic location, and distribution system asset (eg, electrical system substations and circuits, gas lines, water lines). 2. Analytical dashboards that are reached from the performance dashboards and that allow managers and analysts to drill-down into the details of unfavorable variances so that corrective actions can be

Business Intelligence in the Era of Big Data and Cognitive Business

45

quickly identified, evaluated, decided upon, and acted upon. For example, if the performance dashboard identifies that overall system reliability is below the targets projected in justifying the rates, the analytical dashboard allows managers and analysts to drill-down to specific outage events—including location, duration, the assets involved, the priority given to repair, the crew assigned to repair, the extent of the damage, and the predicted time to repair based on established work standards. Armed with this information, corrective action strategies are devised quickly and efficiently. 3. BI applications for demand analysis, predicting demand, predicting the cost of producing power or purchasing power, asset reliability analysis, and asset condition. Achieving customer service and system reliability objectives at the fixed cost assumed during the rate justification process is a complex task. Variances between assumed demand and actual demand induces revenue variances that create pressures on costs due to the need to achieve a return for shareholders. System reliability is impacted by investments in preventative maintenance, the effectiveness of which is impacted by asset conditions and complex decisions as to which assets to maintain and which to run to failure. System reliability is also impacted by outage events and vegetation management processes. Multidimensional analysis, advanced analytics, and predictive analytics are used to understand tradeoffs and take effective asset management and customer service actions. The above BI applications enable utilities to cope with the inherent complexity of their business and to optimize customer service and system reliability at a predetermined cost. They are high-level examples of the kinds of BI capabilities such companies require. Companies that have met these BI requirements can be said to have achieved BI success. For retailers, BI success typically consists of having a comprehensive and specific view of product movement (demand), customer purchasing behavior, product costs, and the impact of price and promotion on product movement—all by store, department, product category, and time of year. Retailers who are advanced with BI leverage point-of-sale (POS) data and syndicated data about product movement to develop a deep understanding of the relationships between product movement, price/promotion, and margins. For retailers who

46

Business Intelligence Strategy and Big Data Analytics

self-distribute and who may manufacture some of their own products, the POS data and syndicated data (demand data) are also used to optimize cost and customer service across the value chain. Further, this same demand data is used with key suppliers to achieve the same goal and avoid stockouts while not holding excess inventory. In this environment, retailers that have achieved a reasonable measure of BI success will have deployed the following BI applications: 1. Executive and store performance dashboards that are updated on a timely basis and that identify the unfavorable performance variances that require immediate management action—typically those variances related to sales, margins, labor utilization, expenses, inventory shrinkage, category performance, and product performance (movement, contribution margin). The variances displayed are typically for the top 10 or so contributors to the unfavorable variances, and the variances are calculated in relation to year-over-year, in relation to an annual operating plan or budget, in relation to updated operating plans or budgets, and by key dimensions such as store, department, subdepartment, product category, and customer segment. 2. Analytical dashboards that are reached from the performance dashboards and that allow managers and analysts to drill-down into the details of unfavorable variances so that corrective actions can be quickly identified, evaluated, decided upon, and acted upon. For example, if the performance dashboard identifies that same-store sales are off on a year-over-year basis, the analytical dashboard allows the analyst to easily identify which store or stores comprise the bulk of the variance and then drill-down to identify departments, categories, and products that are part of the root cause of the variance. Armed with this information, corrective action strategies are devised quickly and efficiently. 3. BI applications for demand analysis, predicting demand under various price and promotion tactics, evaluating product movement trends and category contribution margins, evaluating assortments and product ranges by store, and segmenting customers so that appropriate and personalized rewards, trial offers, and retention offers can be made to optimize customer lifetime value. Many retail businesses are complex because they must stock and sell thousands or tens of thousands of distinct items in a way that results in having the right product and the right price at the time of need for thousands of customers whose needs vary. With fixed shelf space,

Business Intelligence in the Era of Big Data and Cognitive Business

47

companies that tie up space with products that don’t move run the risk suboptimal profits in what is often a tight-margin business. On the other hand, being out-of-stock generates an immediate opportunity cost and may eventually result in customer defections. BI applications for multidimensional analysis, advanced analytics, and predictive analytics are used to understand the fundamental relationship between product, price, promotion, merchandising, and margin so that optimal actions can be taken to retain customers, grow market baskets, and make a reasonable profit. The above BI applications enable retailers to cope with the inherent complexity of their business and to optimize customer retention and company profitability. They are high-level examples of the kinds of BI capabilities such companies require. Companies that have met these BI requirements can be said to have achieved BI success.

2.7 SUMMARY—INDUSTRY VIEWS OF BI SUCCESS The above examples illustrate that what BI success looks like depends on the industry in which a company operates. While the BI tools may be common—dashboards, multidimensional analysis, advanced analytics, predictive analytics, and so forth—the way that the tools are used must be relevant to the industry and the manner in which the company competes in the industry. And while we have focused on five particular industries and the types of BI applications that are relevant for those industries, companies in other industries can leverage well-established business-driven techniques for determining an overall BI vision and identifying which uses of BI are most relevant in their specific cases. These techniques set the stage for BI success and they are subject of the remainder of this book.

2.7.1 Job Function Views of BI Success Our discussion of what BI success looks like in different industries also provided a glimpse of how BI success varies by job function. While the BI needs of people in different job functions within a company are not mutually exclusive, there are definitely BI applications or uses that are job-specific. For example, a plant manager may want a BI application for measuring, managing, and improving plant output by shift and production line. That application would be of limited value to a sales manager. On the other hand, it is not uncommon for sales people,

48

Business Intelligence Strategy and Big Data Analytics

customer service people, and operations people to have a common interest in inventory on hand if the company fulfills orders out of inventory. At a high level, we can generalize that5: • For the Chief Financial Officer (CFO) and financial management professionals, BI success means such things as having a precise and granular understanding of the relationship between operational performance and financial results, having better tools for performance management, having high-quality historical facts at their fingertips for planning, forecasting, and budgeting, and having better information and analytical tools for working capital management. • For the Chief Operating Officer (COO) and operations management professionals, BI success means such things as having precise and granular information available for cost analysis, having analytical tools for monitoring and improving customer service and product quality, and having high-quality historical facts about demand readily available for demand management and capacity planning. • For the Chief Marketing Officer (CMO), sales leaders, and marketing professionals, BI success means such things as having complete information about individual customers to enable better customer segmentation, more precise campaign targeting, improved customer service and customer retention, more timely campaign lift analysis, improved ability to determine customer lifetime value, a better understanding of the price elasticity of demand, improved tools for category management, and tools for performance management. • For the Chief Information Officer (CIO), BI directors, and BI team, BI success means being able to measure BI usage and BI impact, being able to do a better job of meeting the demands of business users, moving beyond being order takers for standard reports, and being able to operate with a solid business case and adequate time and money to be effective in helping improve business performance and profits. The above examples are just a sample of what BI success looks like to the people in companies who are charged with meeting business objectives, delivering profits, and/or meeting competitive challenges. Ultimately, BI success is measured in improved business performance and profitability. Those are the subjects of the rest of this book. 5 Portions of this discussion are taken from Williams S. 5 Barriers to BI success and how to overcome them. Strategic Finance, July 2011.

Business Intelligence in the Era of Big Data and Cognitive Business

49

2.8 RECAP OF SOME KEY POINTS 1. The terms “business intelligence” and “BI” mean different things to different business people. Lack of a common understanding of what BI is and what it can do is an impediment to BI success. 2. There is a lot of hype in the marketplace about BI, big data, and analytics. This confuses business executives, managers, and analysts about the value proposition for BI—which impedes adoption and/or results in ineffective capital investments. 3. Big Data is a combination of traditional business data and new types of “data.” Many of the new types of “data” are actually digital content—like text messages, digital images, music files, etc. This new digital content is referred to as “unstructured data.” To obtain value from unstructured data, it has to be leveraged within a business process that increases revenue, reduces costs, or both. 4. From a BI perspective, big data is simply another source of data and digital content that might be useful for analytical purposes. 5. The primary raw materials for cognitive business are structured and/or unstructured data, computing power, and complex mathematical and statistical methods that are woven into algorithms. 6. The algorithms are designed in conjunction with people who are experts in the relevant business domain, for example, inventory management, insurance fraud detection, operations management, and so forth. 7. Predesigned cognitive business applications will be similar to packaged software, that is, they will be designed by vendors to deliver a standard business technique to as many companies as possible. 8. Custom-designed cognitive business applications will be based on the knowledge about a relevant business process held by people within a given company. For example, a cognitive inventory management application would incorporate knowledge and practices of inventory managers in the given company. 9. Despite the allure of terms like machine learning, artificial intelligence, and mathematical optimization, cognitive business still comes down to applying programmed business logic to data as a means of improving business results. 10. Analytics are not new, though we have better tools for applying them than we did 20 years ago. 11. Historically, analytics have been considered a subset of BI. 12. BI has always been about analysis.

50

Business Intelligence Strategy and Big Data Analytics

13. All forms of BI—including analytics—should be used as appropriate within the core business processes that drive increased revenues, optimized costs, and overall profitability. 14. BI is a general purpose tool that must be applied in different ways for different job functions and industries. For example, a relevant and appropriate BI application for retail grocery store operations improvement will be different from a BI application for customer segmentation for a life insurance company. 15. Because BI must be applied differently in different business contexts, BI success will look different in its specifics depending on the industry, company, and function wherein BI applications are being used.